This browser is no longer supported.
Upgrade to Microsoft Edge to take advantage of the latest features, security updates, and technical support.
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(102.4, 61%, 19%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EWM%3C/text%3E%3C/svg%3E)
Is there a way to book a GPU machine in advance so that the hourly cost is reduced?
Thanks
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(153.6, 16%, 24%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3ERG%3C/text%3E%3C/svg%3E)
Hi Wasif Masood,
Could you please share the VM size you would like to use for booking a GPU machine in advance? This will help us provide the most relevant information and options for your needs.
Thank you.
Hi,
here are needed specs
GPUs = H100x8,
vCPUs >24
Memory 256GB
Hi Wasif Masood,
Thank you for sharing the details. I would like to confirm whether you are specifically looking for the Standard NC40ads H100 v5 VM size, as it differs from the one previously mentioned. Please let us know how you would like to proceed.
I would like to know if you have any machine matching the above mentioned requirements as Standard NC40ads H100 v5 clearly do not.
Hi Wasif Masood,
Azure offers virtual machines that meet your requirements, including the ND H100 v5 series, a flagship addition to Azure’s GPU family. This series features configurations with 8 NVIDIA H100 Tensor Core GPUs, 96 vCPUs, and 1900 GiB of memory, designed for high-end deep learning, generative AI, and HPC workloads.
These VMs provide out-of-the-box GPU acceleration for frameworks like TensorFlow, PyTorch, Caffe, and RAPIDS. Additionally, the scale-out InfiniBand interconnect enables seamless GPU clustering using NVIDIA’s NCCL communication libraries, ensuring optimal performance for AI and HPC applications.
Reference doc, ND-H100-v5 sizes series
Hi Wasif Masood,
Azure GPU resources are typically allocated on-demand, but there are several ways to optimize costs. While GPU Virtual Machines (VMs) have limited availability under Reserved Instances (RI), committing to a reserved instance for general VM families may help reduce expenses. If your workload is flexible, Azure Spot VMs offer significantly lower pricing, though they may be reclaimed based on capacity. Another option is Azure Savings Plans, which allow you to commit to a spending amount over one or three years to lower GPU costs. If ensuring availability is a priority, Azure Capacity Reservations enable you to secure compute capacity in advance, though they do not provide cost savings. Please let us know if you need further guidance in selecting the best option for your requirements.
Azure for Startups | Microsoft Learn
Please reach out to us if you have any other queries.
If the information is helpful, please Accept Answer & Upvote so that it would be helpful to other community members.
Glad to hear that your issue got resolved. Please don't hesitate to reach out to us if you have any further queries.
Thank you